Author: AILesson9 min setupTested with:ChatGPTReviewed: 2026-08-28
Quick answer
Turn verified product behavior into safe, scannable steps with prerequisites, outcomes, errors, and escalation. Provide: Verified product behavior, User task and evidence, Publishing and support rules. Expected result: A task-complete help article with tested steps, accessibility, troubleshooting boundaries, and maintenance metadata.
1
Add your context
Your text stays in this browser. AILesson Prompts does not send it to a model or server.
2
Your prompt
Unfilled fields remain visible as placeholders, so you can still copy and edit the prompt
Draft a self-service help-center article using only verified product behavior.
Product/version, platforms, roles, prerequisites, UI labels, steps, results, errors, limitations, and sources:
[behavior]
User goal, starting state, support evidence, language, terminology, accessibility, risk, and success evidence:
[users]
Channel, style, screenshots, privacy/security, permissions, escalation, owner, review date, and localization:
[publishing]
Confirm that the supplied evidence supports the complete task. Do not invent menus, buttons, defaults, screenshots, keyboard behavior, availability, causes, recovery, or policy. If platforms/roles differ, branch explicitly. Separate prerequisite, action, expected visible result, verification, reversible warning, error path, and human-escalation trigger. Never ask users to expose passwords, tokens, full payment details, personal records, or restricted logs. Security/safety incidents leave ordinary troubleshooting when policy says so.
Write search-oriented title, one-sentence outcome, applies-to metadata, before-you-start checklist, numbered steps with one action each, expected result after meaningful steps, verify-success section, common issues, escalation/contact path, related articles, and last-reviewed metadata. Use exact UI text, plain language, meaningful link labels, accessible text alternatives for necessary images, and localization notes. Provide screenshot plan rather than fake images, a source-to-step ledger, task-test script for normal/permission/error paths, and blockers that prevent publication.
Private by defaultPrompt assembly happens locally in your browser. Avoid placing confidential information into any AI service unless your organization allows it.
From input to outcome
A worked example
See how concrete context turns this recipe into a usable result
Actual input
Verified product behavior
ClearLedger web app 8.14, desktop browsers only. Editors can import CSV; reviewers/admins can review but reviewer role cannot import. Source P14 verified 27 Aug 2026: Projects > select project > Imports > New import; choose CSV up to 10 MB; first row must contain unique nonblank headers; select dictionary; Map fields; Run validation; invalid rows appear under Issues and can be downloaded as CSV. Import does not auto-correct source. Valid rows remain Draft until reviewer chooses Approve import. Error E17 'Duplicate or blank header'; E22 'File exceeds 10 MB'. No mobile import. Security policy S3: unexpected account-access/reset events go to security support, not import troubleshooting.
User task and evidence
Goal: branch reporting editor imports monthly CSV and identifies invalid category labels before review. Starting state: signed in, project access granted, dictionary already created. Support contacts show users expect automatic correction and reviewers look for import button. English and Chinese UI labels are approved. Need keyboard-accessible instructions, but keyboard sequence for native file picker varies by OS and is not documented.
Publishing and support rules
Public bilingual help center, plain language. Use screenshots only from synthetic demo tenant; no account names/data. Support route is in-app Help > Contact support; security route is Help > Security concern. Content owner Mina; product reviewer Ravi. Publish only after desktop Chrome/Edge/Firefox tests. Review date 30 Nov 2026. Screenshot captions/alt text required.
Example output
Title: Import and validate a CSV in ClearLedger
Outcome: Editors can upload a monthly CSV, map it to an existing dictionary, and download invalid rows for correction before review. Applies to ClearLedger web 8.14 on desktop Chrome, Edge, and Firefox; Editor role only. Reviewers cannot start imports. Mobile import is unavailable.
Before you start: sign in; confirm Editor access to the project; prepare a CSV ≤10 MB whose first row has unique, nonblank headers; confirm the dictionary exists. Keep a safe copy of the source because ClearLedger does not auto-correct it.
Select Projects, then open the project. 2. Select Imports > New import. 3. Choose the CSV. Expected: the file name appears for mapping. 4. Select the existing dictionary. 5. Under Map fields, match each CSV header to the intended dictionary field. 6. Select Run validation. Expected: valid rows remain in Draft; invalid rows appear under Issues. 7. Open Issues, then download the issues CSV. Correct the original source outside ClearLedger and start a new import as required by your workflow. The product does not correct rows automatically. 8. A Reviewer verifies the Draft and selects Approve import; this is a separate role action.
Verify: status shows approved only after reviewer action; unresolved invalid rows are not silently corrected. E17: inspect the first row for blank/duplicate headers. E22: use a file ≤10 MB. Other errors: Contact support and share error code, browser/version, and synthetic-safe steps—not the source file unless the approved channel requests it. Unexpected reset/access events: stop and use Help > Security concern.
Screenshot plan: synthetic Imports page, mapping, Issues, Draft/Approve; alt text names controls/result, no data. Test Editor normal/E17/E22, Reviewer permission, three browsers. Keyboard file-picker sequence remains undocumented: publication blocker for any claimed exact OS keystrokes, so instruct users to use their operating system's standard picker without inventing keys. Ledger: steps/errors/roles P14; security S3. Owners Mina/Ravi; review 30 Nov 2026.
Why this works
1
Expected results after key actions let users detect divergence before later steps make the problem harder to diagnose
2
A source-to-step ledger keeps UI instructions maintainable when product versions change
Check the result
Is every UI label, action, result, error, role, and platform branch supported by current sources
Can a user verify completion and recover or escalate safely without exposing restricted data
Are accessibility, screenshots, localization, test paths, ownership, and review date publication-ready
Use it with confidence
Frequently asked questions
Practical answers about when to use this recipe, what to provide, and where human review still matters
What should I prepare before using “Draft a self-service help center article”?
For “Draft a self-service help center article,” prepare Verified product behavior, User task and evidence, and Publishing and support rules. Replace placeholders only with information you can verify. If a detail is unknown, preserve that uncertainty explicitly instead of asking the model to infer it.
When is the “Draft a self-service help center article” result not ready to use?
The result is not ready if it does not yet deliver the stated outcome—A task-complete help article with tested steps, accessibility, troubleshooting boundaries, and maintenance metadata—from the supplied evidence, or if it relies on unresolved assumptions, missing approvals, or invented details. Use the checks as release gates: revise the source inputs or assign a named, authorized reviewer instead of polishing an unsupported output.
Which AI tools have recorded tests for “Draft a self-service help center article”?
The published test record for “Draft a self-service help center article” lists ChatGPT as of 2026-08-28. This confirms recorded runs, not guaranteed compatibility or identical results in later product versions. For another tool or version, keep every constraint visible and repeat the result checks before use.
ChatGPT